Please use this identifier to cite or link to this item:

http://hdl.handle.net/10609/98366
Title: Modelització de dades longitudinals amb efectes aleatoris de intercepció/pendent i presència de dades perdudes en el context dels estudis d'estabilitat
Author: Bernad Martin, Francesc
Director: Pérez Álvarez, Nuria
Tutor: Ventura Royo, Carles  
Keywords: stability studies
mixed models
missing data
Issue Date: Jun-2019
Publisher: Universitat Oberta de Catalunya (UOC)
Abstract: The purpose of this work is to analyze the application of mixed models and models with presence of missing data in the context of statistical analysis in studies of stability of biological products or similar. The purpose is to evaluate the applicability of these two points to explain and predict stability studies. The applied methodology has been the search for bibliographic information in order to perform a theoretical analysis taking into account the specific conditions of a usual stability study and using this information, to be able to generate a simulation model for the application of mixed models and a model of simulation to evaluate the presence of missing data and some methods of data imputation. The results of the mixed models simulations show the most suitable models to apply according to the data behavior, presenting the advantages of applying random effects and the usefulness of the application of variance modulation functions. The results of the models with missing data simulations show the biases that can occur with the presence of missing data combined with the application of mixed models, and the efficiency or deviation of two simple imputation methods. The practical exercice results shows the mixed models analysis reproducibility in a real case and how, differences aside, some of the results seen in the previous simulation are reproduced.
Language: Catalan
URI: http://hdl.handle.net/10609/98366
Appears in Collections:Bachelor thesis, research projects, etc.

Share:
Export:

This item is licensed under a Creative Commons License Creative Commons